Methodology

ElonFantasy is a public accountability platform that scans posts on X (formerly Twitter) from high-profile politicians, tech leaders, and media figures. It uses AI to detect contradictions and flip-flops, then scores each profile on consistency, accuracy, and value alignment. The goal is radical transparency: anyone can look up an influential person and see a documented, sourced record of what they said versus what they did.

Editorial Disclaimer

Truth Scores represent AI-assisted editorial judgment, not objective fact. Our system uses large language models to detect semantic contradictions and assess consistency. These assessments are probabilistic and may contain errors. We encourage users to review the source material and draw their own conclusions.

Scoring System

Each tracked figure receives a composite Truth Score from 0 to 100, calculated from three weighted sub-scores:

Sub-ScoreWeightMeasures
Consistency40%Semantic similarity between past and present statements on the same topic
Accuracy35%Factual claims cross-referenced against public records and fact-checking databases
Value Alignment25%Comparison of stated values/principles against observed behavior and policy decisions

Truth Score = (Consistency x 0.4) + (Accuracy x 0.35) + (Alignment x 0.25)

Verdict Tiers

85-100
Truth Anchored

Exceptionally consistent; actions closely match stated values

70-84
High Integrity

Rarely contradicts; generally reliable and principled

55-69
Signal with Noise

Mostly consistent but with notable inconsistencies

40-54
Narrative Driven

Frequent flip-flops; positions shift with convenience

25-39
Low Fidelity

Documented pattern of contradicting prior statements

0-24
Reality Distortion

Persistent, severe contradictions across core stated values

Contradiction Detection

Our AI pipeline identifies contradictions through a multi-step process:

  1. Posts are collected from public X accounts via the X API
  2. All posts are fed to Kimi K2 for semantic contradiction analysis
  3. The AI identifies post pairs with genuinely incompatible positions on the same topic
  4. A valid contradiction requires: the same subject matter, meaningfully opposing positions, and cannot be explained by new information or an acknowledged change of mind
  5. Each contradiction is assigned a severity score (1 = minor, 2 = clear, 3 = major)
  6. Sub-scores are calculated based on contradiction volume, severity, and factual accuracy

Data Sources

All content is sourced from public posts on X (Twitter). We do not access or display private messages, deleted posts, or any non-public information. Engagement metrics (likes, retweets) are used for weighting high-signal posts but do not directly affect scores.

Dispute Process

If a public figure or their representative believes a contradiction has been incorrectly flagged, they can submit a dispute through our contact form. Disputes are reviewed by our editorial team and AI reassessment. Legitimate disputes (e.g., acknowledged position changes with new information) will result in the contradiction being updated or removed.